資源描述:
《基于多agent的生產(chǎn)計(jì)劃與調(diào)度系統(tǒng)研究與開發(fā)》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、基于多Agent的生產(chǎn)計(jì)劃與調(diào)度系統(tǒng)研究與開發(fā)摘要多Agent是多學(xué)科相融合、具有很高實(shí)用價(jià)值的研究領(lǐng)域,是當(dāng)前人工智能領(lǐng)域的研究熱點(diǎn)。生產(chǎn)調(diào)度位于CIMS體系結(jié)構(gòu)的中間層,是實(shí)施CIMS的關(guān)鍵。在實(shí)際制造系統(tǒng)開放的、動(dòng)態(tài)的環(huán)境r,生產(chǎn)調(diào)度問題不僅是具有NP難度的組合優(yōu)化問題,而且呈現(xiàn)極強(qiáng)的動(dòng)態(tài)性,借助多Agent的自治性和合作能力為解決動(dòng)態(tài)的、復(fù)雜的調(diào)度問題提供了可能。本文系統(tǒng)地闡述了多Agent理論及其在生產(chǎn)調(diào)度中的應(yīng)用,通過多Agent技術(shù),把計(jì)劃分配與任務(wù)調(diào)度有機(jī)的結(jié)合起來,實(shí)現(xiàn)整個(gè)系統(tǒng)的計(jì)算機(jī)集成。具體研究工作包括以下幾個(gè)方面:1.通過對(duì)目前較為流行的合同網(wǎng)協(xié)議的分析,指出,
2、合同嗍協(xié)議存在通訊量大、協(xié)商效率低等缺點(diǎn),并提出了一種改進(jìn)的合同網(wǎng)模型用以解決分布式車間的計(jì)劃分配問題。2.把蟻群算法應(yīng)用于Job.shop調(diào)度問題。提出了一種基于工序的螞蟻遍歷方法,并對(duì)揮發(fā)系數(shù)引入了一個(gè)自適應(yīng)過程,通過解的特征自動(dòng)的來調(diào)整揮發(fā)系數(shù)。3.通過把蟻群算法與強(qiáng)化學(xué)習(xí)相結(jié)合,提m了一種基于自適麻Agent的車間調(diào)度方法。當(dāng)生產(chǎn)環(huán)境發(fā)生變化時(shí),螞蟻會(huì)根據(jù)歷史獎(jiǎng)勵(lì)和立即獎(jiǎng)勵(lì)情況進(jìn)行決策,實(shí)現(xiàn)任務(wù)在機(jī)器資源上的分配。l4.總結(jié)了前人的研究成果,將遺傳算法和神經(jīng)網(wǎng)絡(luò)等算法與本文所研究的蟻群算法、強(qiáng)化學(xué)習(xí)算法一并封裝成‘個(gè)調(diào)度算法席,通過改進(jìn)的合同網(wǎng)協(xié)議把整個(gè)系統(tǒng)連接起來,完成生產(chǎn)計(jì)
3、劃與調(diào)皮系統(tǒng)的開發(fā)。關(guān)鍵詞:智能體,生產(chǎn)調(diào)度,蟻群算法,強(qiáng)化學(xué)習(xí)IlRESEARCHANDDEVELOP~ⅡNT0FPRODUCTIONPLANNINGANDSCHEDULINGSYSTEMBASED0NMUITI.AGENTABSTRACTMulti··Agenttechnologyismulti·-subjectcrossedresearchfieldanditsapplicationshowsthehighvalue.Itisthehotspotintherecentstudyonartificialintelligence.AsthemiddlelayerintheCMIS,p
4、roductionschedulingisthekeyoftheCIMS.Undertheopenanddynamicenvironmentofrealmanufacturesystem,itistheNPhardcombinatoriaoptimizationproblemanditalsobehavesgreatlydynamiccharacteristicsWiththeautonomousandcooperativeabilityoftheMulti—Agent,itispossibletosolvethecomplexanddynamicschedulingproblemIn
5、thispaper,theMulti—Agenttechnologyanditsapplicationintheproductionschedulingwereintroduced.TheplansdistributingandassignmentsschedulingwereintegratedintothewholesystembyMulti-Agenttechnology.Themainresearchworkisdescribedasfollowing1.Thepopularcontractnetprotocolwasanalyzed.Thenthedisadvantagesw
6、erepresented,whichwerecrowdingcommunicationandⅡIthelownegotiatingefficiency.Finallyanewimprovedcontractnetmodelwasproposedtosolveproblemofdistributedshopplanning2.Antcolonyalgorithmwasusedtosolvejob-shopschedulingproblem.Anewwayofantcrawlingwasproposed.Intermofcharacteristicofsolutions,anadaptiv
7、eadjustmentprocessofthqvolatilitycoefficientwasintroduced.3.Combiningantcolonyalgorithmandreinforcementlearning,anewjob—shopschedulingalgorithmbasedonanadaptiveagentwasproposed.Whenproductionenvironmentchanged,theartificiala